Tumor suppression by a proapoptotic calcium-activated chloride channel in mammary epithelium.

Little is known of the roles played by ion channels in cancer. Here we describe a pair of closely related calcium-activated chloride channels whose differential regulation in normal, apoptotic, and transformed mouse cells suggests that channel function is proapoptotic and antineoplastic. While mCLCA1 predominates over mCLCA2 under normal physiological conditions, this relationship is reversed by apoptotic stress both in developing mammary gland and in cultured HC11 mammary epithelial cells. Consistent with an apoptosis-promoting role, splicing of mCLCA2 is disrupted in apoptosis-resistant tumor cell lines and in HC11 cells selected for resistance to detachment-induced apoptosis (anoikis). Unexpectedly, mCLCA1 message is also down-regulated in these cells by at least 30-fold. These results suggest that both genes antagonize survival of mammary tumor cells by sensitizing them to anoikis. When MCF7 or HEK293 tumor cells were transfected with plasmids encoding either mCLCA1 or mCLCA2, colony formation was greatly reduced relative to a vector-transfected control, demonstrating that calcium-sensitive chloride channel (CLCA) expression is deleterious to tumor cell survival. Furthermore, mammary epithelial cells overexpressing mCLCA2 had twice the rate of apoptosis of normal cells when subjected to serum starvation and formed multinuclear giants at a high frequency in normal culture, suggesting that mCLCA2 can promote either apoptosis or senescence.

Little is known of the roles played by ion channels in cancer. Here we describe a pair of closely related calciumactivated chloride channels whose differential regulation in normal, apoptotic, and transformed mouse cells suggests that channel function is proapoptotic and antineoplastic. While mCLCA1 predominates over mCLCA2 under normal physiological conditions, this relationship is reversed by apoptotic stress both in developing mammary gland and in cultured HC11 mammary epithelial cells. Consistent with an apoptosis-promoting role, splicing of mCLCA2 is disrupted in apoptosis-resistant tumor cell lines and in HC11 cells selected for resistance to detachment-induced apoptosis (anoikis). Unexpectedly, mCLCA1 message is also down-regulated in these cells by at least 30-fold. These results suggest that both genes antagonize survival of mammary tumor cells by sensitizing them to anoikis. When MCF7 or HEK293 tumor cells were transfected with plasmids encoding either mCLCA1 or mCLCA2, colony formation was greatly reduced relative to a vector-transfected control, demonstrating that calcium-sensitive chloride channel (CLCA) expression is deleterious to tumor cell survival. Furthermore, mammary epithelial cells overexpressing mCLCA2 had twice the rate of apoptosis of normal cells when subjected to serum starvation and formed multinuclear giants at a high frequency in normal culture, suggesting that mCLCA2 can promote either apoptosis or senescence.
The CLCA 1 family of calcium-activated chloride channels comprises about a dozen members thus far expressed in a variety of organ and tissue types in mammals (1)(2)(3)(4)(5)(6)(7)(8). Although the family is deeply divergent, some members sharing as little as 36% identity, all share several features. The CLCA precursor is about 900 amino acids long with one proteolytic cleavage site following the amino-terminal signal sequence and another around residue 600. The latter results in a 90-kDa product with perhaps four transmembrane segments and a 30 -40-kDa product that may be found in close association with the first (1,3,4,6). Another signatory feature is a symmetrical multiple cys-teine motif, CX 12 CX 4 CX 4 CX 12 C, in the amino-terminal tail. Although a structure-function analysis has not been performed on a CLCA family member, potential phosphorylation sites for calmodulin kinase II and protein kinase C are consistent with calcium regulation. Indeed, all CLCA proteins tested thus far have been shown to be outwardly rectified chloride channels that are activated by calcium (2)(3)(4)6). Recently, a human CLCA homolog, hCLCA2, was found to be highly expressed in epithelium of mammary ducts and alveoli but down-regulated in mammary tumors and tumor cell lines (9). Ectopic expression of hCLCA2 in breast cancer cells inhibited subcutaneous tumor formation in nude mice by an unknown mechanism, suggesting that hCLCA2 is a tumor suppressor.
We described previously a calcium-activated chloride channel derived from mouse lung, mCLCA1 (3). Hybridization studies indicated widespread expression in secretory tissues, including mammary ductal and alveolar epithelium (5). To determine whether a signal from a mammary gland represented mCLCA1 or a closely related homolog, we probed lactating mammary glands by degenerate PCR. The predominant product was a homolog of mCLCA1, referred to here as mCLCA2. High-level expression of mCLCA2 correlated closely with the onset of apoptosis, both in vivo and in vitro. A role for chloride channels in apoptosis induction has been suggested previously. In lymphocytes, an outwardly rectified chloride channel has been shown to be rapidly activated in response to apoptogens, and channel inhibitors blocked apoptosis (10). Similarly, mutations in the cystic fibrosis transmembrane regulator, CFTR, were found to inhibit apoptosis in C127 cells (11). We reasoned that if CLCA channels played a similar role in mammary epithelial cells, then the mCLCA2 gene should be down-regulated in cells that have become apoptosis-resistant, e.g. tumor cell lines. To find whether selection for apoptosis resistance would suffice to evoke down-regulation of CLCA genes, we selected in vitro for resistance to anoikis, the form of apoptosis driven by cell detachment in the absence of growth and survival factors, and examined CLCA expression (12,13). Finally, we asked whether re-expression of CLCA genes would be tolerated by tumor cell lines.
Technologies, Inc.). A 2.7-kilobase product was amplified by PCR (93°C, 30 min; 55°C, 45 min; 72°C, 2Ј; 35 cycles) using primers  corresponding to the 5Ј-and 3Ј-ends of the mCLCA1 open reading frame  (5Ј-TGGAGCAGTGCGACCATGGTGCCA-3Ј and 5Ј-TAGTTTAAAATA-GAGTTAAAAATCACAG-3Ј) and the high fidelity Pwo DNA polymerase. This product was sequenced directly then cloned into pGEM-T (Promega). Of five clones, one corresponded in sequence to mCLCA1. The other clones and the directly sequenced PCR product contained a distinct sequence that was called mCLCA2. Differences between the two open reading frames were confirmed by amplifying and sequencing smaller intervals of the cDNAs from virgin or lactating glands from two different reverse-transcription reactions and commercially prepared cDNAs (RapidScan, Origene).
Relative Expression of mCLCA1 and mCLCA2 in Mammary cDNA-cDNA from mammary glands of different developmental stages (RapidScan kit, Origene) was analyzed by PCR with a primer pair designed to symmetrically bracket an EcoRI site present in the mCLCA2 cDNA at nucleotide 866 but absent from the mCLCA1 cDNA. Primers were 5Ј-ACTCGAAGACACGGCTGTATGAAC-3Ј and 5Ј-CT-GTCAAATGTGACTAATCCAAC-3Ј, starting at nucleotides 650 and 1075, respectively. Thus, cleavage of the PCR product with EcoRI and electrophoresis resulted in a 425-base pair band for mCLCA1 and a doublet of 212 and 213 base pairs for mCLCA2. This approach to relative quantitation was designed to avoid problems due to differences in amplification efficiency observed with gene-specific primer pairs. PCR was for 35 cycles (93°C, 30 min; 55°C, 45 min; 72°C, 45 min) using Taq DNA polymerase (Life Technologies, Inc.). The cycle number was selected to be within the linear phase of PCR with respect to mCLCA1 product/template ratio. For amplification of the ␤-actin control, template was diluted 10-fold, and the cycle number was 33. Primers for ␤-actin were 5Ј-TGGTGGGAATGGGTCAGA-3Ј and 5Ј-AGGGAGGAAGAGGATGCG-3Ј.
Induction of Gene Expression by Lactogenic Hormones-10-cm dishes of HC11 cells were maintained at confluency for 4 days in CM, then primed by incubation in RPMI/2% FBS/insulin (5 g/ml) for 2 days before inducing with prolactin (5 g/ml) and dexamethasone (1 g/ml). After 4 days, the medium in the last dish was replaced with RPMI without supplements, and cells were harvested after a further 2 days. RNA was extracted with Trizol, and 1 g of total RNA was reversetranscribed with Superscript reverse transcriptase (Life Technologies, Inc.). Initially, serial dilutions of these reactions were amplified 35 cycles with primers specific for mCLCA1 and mCLCA2, ␤-casein, or GAPDH. Dilutions were selected that produced a linear response with respect to an added template for each primer pair. Primers for ␤-casein were 5Ј-GTGGCCCTTGCTCTTGCAAG-3Ј and 5Ј-AGTCTGAGG-AAAAGCCTGAAC-3Ј.
Amplification of mCLCA1 and mCLCA2 from Multi-tissue Array-A multi-tissue array of cDNAs normalized to ␤-actin (RapidScan, Origene) was probed by PCR (35 cycles) as described above for mammary cDNA.
Selection of Apoptosis-resistant Cells and Tumorigenicity Assay-HC11 cells were grown to confluency, then detached with trypsin, diluted into RPMI, and incubated 3-4 days until most of the cells were dead. Surviving cells were allowed to recover in 10% serum and to return to confluency. Timing was adjusted to allow about 5-10% survival, and the cycle was repeated 10 times. A control dish was cultured continuously in 10% serum during this period. RNA was extracted, and the expression levels of mCLCA1 and mCLCA2 compared with those of the untreated control by RT-PCR, taking care that all dishes were equally confluent for the same period of time before harvest. The resulting cell lines SS1 and SS2 were tested for their tumorigenicity by injecting 10 6 cells subcutaneously at the right gluteus of three BALB/c mice each. Sets of three control mice were injected similarly with either non-selected HC11 or JC tumor cells.
Analysis of Tumor and Apoptosis-resistant Cell Lines-Cells were grown to confluency in Dulbecco's modified Eagle's medium plus 10% FBS. RT-PCR was performed as described above (650 -1075 interval) except that PCR was limited to 31 cycles. Serial dilutions of HC11 cDNA were included for relative quantitation. For detection of splice variants, cDNA prepared from HC11 or tumor cell lines was probed systematically by PCR with a number of primer sets derived from the mCLCA1 and mCLCA2 cDNAs, designed to amplify either the entire open reading frame or smaller segments. When aberrant-sized products were detected, their authenticity was first confirmed by Southern hybridization or nested PCR before determining their sequences.
Transfection and Selection-The mCLCA2 cDNA was excised from pGEM-T as a NotI/SalI fragment and transferred to pcDNA3.1Zeo. The mCLCA1 cDNA was excised from pTet-Splice (3) and inserted into pcDNA3.1Zeo as an EcoRI/SalI fragment. The resulting plasmids were transfected into low passage HEK293 cells with LipofectAMINE Plus, and zeocin selection (400 g/ml) was applied. Co-transfection with pEGFP-N1 (CLONTECH) provided a control for transfection frequency, which did not vary significantly between DNA preparations. After death of the untransfected cells (5 days), surviving cells were transferred to a 12-well plate and allowed to form colonies under continued selection. MCF-7 and HC11 cells were transfected using Lipo-fectAMINE 2000 and selected similarly. To visualize and quantify colony survival, cells were stained after 3 weeks with the basic dye crystal violet as described by Carnero et al. (15). After washing the cells extensively with water, cells were photographed, and then stain was eluted with 10% acetic acid and quantified spectrophotometrically at 595 nm as a measure of surviving cell number.
Retroviral Infection and Apoptosis Assay-For retroviral infection, the mCLCA2 cDNA was inserted into the Pinco vector at the BamHI site, and viral supernatants were generated from Phoenix E cells as described (16,17). This vector carries a green fluorescent protein (GFP) marker. HC11 cells were infected overnight as nearly confluent monolayers. After 1 week, cells were trypsinized and diluted into serum-free medium. After 3 days, aliquots of cells were removed, and the DNAspecific fluorescent dyes Hoechst 33342 and propidium iodide (PI) were used to distinguish between apoptotic and necrotic cells. While Hoechst is membrane-permeable and stains all nuclei, PI stains only cells that have lost membrane integrity, i.e. necrotic cells (18). Cells were considered apoptotic if nuclei were PI-negative and clearly disintegrating into apoptotic bodies. About 500 GFP-positive cells were counted for each measurement, and results from two dishes were averaged.

RESULTS
Identification and Cloning of mCLCA2-CLCA homologs were amplified from mammary glands of lactating mice as described in "Material and Methods." Cloning of the products revealed two distinct sequences, one identical to mCLCA1 and the other encoding a protein 95% identical to mCLCA1, mCLCA2 (Fig. 1). PCR error was ruled out by using a high fidelity polymerase, directly sequencing PCR products before cloning, and sequencing multiple clones. Amino acid differences between the homologs are generally scattered and do not affect known features of the CLCA family such as phosphorylation sites or the symmetrical multi-cysteine region (1, 3). One potential glycosylation site is lost from mCLCA2 at amino acid 687 and another is gained at amino acid 840. Expression of the cDNA in HEK293 cells revealed similar protein precursor and processing products as observed for mCLCA1 and other CLCA family members (data not shown; Ref. 3). Recently, a similar cDNA sequence from lactating mammary gland has been reported (19). The sequence of mCLCA2 in Fig. 1 differs at six amino acid positions but probably represents the same cDNA because RT-PCR with primers that would recognize either cDNA revealed only the sequence shown here in mammary gland (not shown).
Differential Expression of mCLCA1 and mCLCA2 during Mammary Development-To assess whether mCLCA1 and mCLCA2 were expressed similarly in mammary development, cDNA from virgin, pregnant, lactating, or involuting mouse mammary glands were analyzed by PCR as described in "Materials and Methods" (Fig. 2A). Expression of mCLCA1 and mCLCA2 was nearly reciprocal. While mCLCA1 levels were similar in virgin, pregnant, and lactating gland but abated during involution, mCLCA2 mRNA was low in virgin gland but greatly increased during lactation and remained high in involuting gland ( Fig. 2A).
Induction of mCLCA2 by Apoptotic Stimuli But Not by Lactogenic Hormones-The induction of mCLCA2 and repression of mCLCA1 might be due either to changes in lactogenic hormone levels during the onset and cessation of lactation or to loss of growth factors and disruption of the extracellular matrix during mammary reorganization and involution (13,20). These two possibilities were tested independently in vitro. The immortalized mammary epithelial cell line HC11 is known to induce transcription of ␤-casein in response to the lactogenic hormones prolactin, insulin, and hydrocortisone and to undergo apoptosis when deprived of serum or anchorage (13,14). Accordingly, when HC11 cells were grown to confluency in CM then transferred to medium without serum, over the course of 48 h cells began to detach from the plate and showed signs of apoptosis (13). During this time, expression of mCLCA1 abated while that of mCLCA2 was strongly induced (Fig. 2B). Transcription of Bax, a marker for apoptosis, also increased significantly, while GAPDH, an internal standard, showed no change. Similarly, anchorage deprivation by growth on poly-(HEMA)-treated dishes elicited an increase in mCLCA2 transcript levels (Fig. 2C). In contrast, treatment of HC11 with lactogenic hormones did not induce expression of mCLCA2, although mCLCA1 was induced in parallel with ␤-casein, reaching its peak 2 days after hormone addition and declining after hormone and serum withdrawal (Fig. 2D). By 2 days after withdrawal of hormones and growth factors, neither mCLCA1 nor mCLCA2 transcript was detected, in agreement with Fig.  2A. Thus, increase in mCLCA2 expression observed during mammary development in vivo is triggered by loss of anchorage and growth factors and not by hormones.
Increased Levels of mCLCA2 mRNA in Tissues with Higher Rates of Cell Turnover-The results above show that, at least in mammary epithelial cells, mCLCA1 and mCLCA2 are expressed in the same cell type but under differing physiological circumstances. mCLCA1 normally predominates, whereas apoptotic stress induces mCLCA2. To assess whether this relationship prevailed in other tissues, an array of cDNAs from various organs and developmental stages was probed by RT-PCR (Fig.  3). As in mammary, expression of both mCLCA1 and mCLCA2 was observed in most of the tissues examined; however, mCLCA1 usually predominated. The exceptions were thymus, small intestine, and skin, all of which are typified by high rates of cell turnover by apoptosis (21)(22)(23). Similarly, high expression of mCLCA2 during embryonic development may reflect reorganization by apoptotic cell death (24).
Down-regulation of mCLCA1 and mCLCA2 in Transformed and Apoptosis-resistant Cells-These results, together with the demonstrations that Cl Ϫ channels are required for apoptosis in lymphocytes and that CFTR can promote apoptosis in mammary epithelial cells (10,11), suggested that mCLCA2 might be a component of an apoptosis-promoting pathway. Since one of the FIG. 1. Comparison of mCLCA2 amino acid sequence with that of mCLCA1. The proteins are 95% identical with broadly dispersed differences that do not affect known sites for glycosylation (asterisks) or proteolytic processing (diamond). However, seven substitutions are concentrated within the symmetrical multicysteine region (underlined). Bars, regions missing from principal splice variants expressed in tumor cell lines CSML-0 (diagonal filled) and JC (shaded). early requisite steps in tumorigenesis is inactivation of apoptotic pathways by mutation, we examined expression of mCLCA1 and mCLCA2 in two cell lines derived from spontaneous mammary adenocarcinomas, JC (25) and CSML-0 (26), with the expectation that mCLCA2 expression would be reduced especially when cells were starved. In both cell lines, net CLCA expression was indeed reduced by at least 30-fold relative to non-transformed HC11 cells even when grown in the presence of serum (Fig. 4A). Surprisingly, further analysis revealed that mCLCA1 transcript had been reduced, but not that of mCLCA2 (Fig. 4B). However, when amplification of the entire mCLCA2 open reading frame was attempted, the tumor cells proved to express only truncated splice variants incapable of encoding full-length mCLCA2 protein ( Figs. 1 and 4C). For example, JC cells expressed transcripts lacking coding sequences for amino acids 360 -730 or 654 -708, while the principal transcript detected in CSML-0 cells lacked amino acids 75-383 (Fig. 1). The PCR products clearly represented splice variants rather than genomic deletions because the reading frames were always conserved, deletions generally corresponded to intron/exon junctions mapped in the hCLCA1 gene (4), and occasionally different cDNAs from the same cell line lacked different intervals. For example, the right-hand panel of Fig. 4C shows the products obtained from amplification of a 1.3-kilobase region containing the conserved proteolytic processing site. JC cells produced one transcript that deleted the site and one that did not. The upper band in the JC and CSML lanes was not characterized. No evidence of mCLCA1 splice variants was found, suggesting that mCLCA1 expression is disrupted at the promoter level in tumor cells.
To address the question whether loss of mCLCA1 and mCLCA2 expression is an obligatory step in the acquisition of FIG. 2. Reciprocal regulation of mCLCA1 and mCLCA2 in response to stress and lactogenic hormones. A, developing mammary gland. cDNA was generated from virgin (V), pregnant (P), lactating (L), or involuting (I) mammary gland and subjected to PCR using primers that recognize both mCLCA1 and mCLCA2 spanning base pairs 650 -1075. The product was cleaved with EcoRI, taking advantage of an EcoRI site found at the center of this interval in mCLCA2 but not mCLCA1. Thus, the upper band represents mCLCA1 and the lower doublet, mCLCA2. ␤-actin was amplified as an internal control. B, serum starvation. HC11 mammary epithelial cells were grown to confluency in RPMI/10% FBS, then washed with and incubated in serumfree medium for the indicated number of hours. C, anchorage deprivation. Cells were grown to confluency as in B, and then plated on plastic (Pl) or dishes treated with poly(HEMA) (Ph) to prevent attachment. Cells were harvested 30 h later. D, lactogenesis. Cells were grown to confluency as in B and treated with lactogenic hormones (prolactin, insulin, and hydrocortisone) for 0 to 6 days followed by hormone withdrawal for 2 more days (WD). Induction of mCLCA1 by hormones was observed in three separate experiments. In B-D, RNA was extracted, reverse-transcribed, and then analyzed as in A. Bax, casein, and ␤-actin or GAPDH were included as controls, respectively, for apoptosis, lactogenesis, and RNA amount.

FIG. 3. Organ and tissue distribution of mCLCA2 expression.
cDNAs generated from various mouse organs or whole embryonic mouse were subjected to PCR as described in Fig. 2.

FIG. 4. RT-PCR analysis of mCLCA1 and mCLCA2 expression in normal and transformed mouse cells.
A, semi-quantitative PCR by template dilution shows that net CLCA expression is about 30-fold lower in transformed (JC and CSML-0) and apoptosis-resistant (SS1 and SS2) cells than in normal HC11 mammary epithelial cells. The dilution factor is indicated above each HC11 lane. PCR was performed with the same primer set as in Figs. 2 and 3, but the product was not cleaved. B, samples shown in A were amplified a further 10 cycles and then the product was cleaved with EcoRI to determine relative expression of mCLCA1 and mCLCA2. C, aberrant splicing of mCLCA2 in tumor cells. Three intervals of mCLCA2, 150 -2700 (left panel), 18 -2595 (center), and 1395-2595 (right), were amplified from randomly primed cDNAs of mammary epithelial (HC11) or mammary adenocarcinoma (JC, CSML) cells, and products were sequenced. Deleted regions are indicated in Fig. 1. For simplicity, only the most prevalent splice variants are shown. Plasmid clones of mCLCA1 or mCLCA2 served as controls for primer specificity (not shown). resistance to apoptosis, HC11 cells were subjected to numerous cycles of apoptotic stress and re-growth by withdrawing serum and anchorage for 3 days followed by serum addition and return to confluency. While initial cell loss by apoptosis was Ͼ90%, with each subsequent round of starvation fewer cells rounded up, resulting eventually in a fully anoikis-resistant population. Analysis of the cells by RT-PCR showed that, like the tumor cell lines, the selected cell lines SS1 and SS2 had lost expression of mCLCA1 (Fig. 4, A and B). The change in expression was stably inherited over 2 months of serial passage in 10% FBS. This suggests that apoptotic pressure during the course of tumor development selects for cells that have lost CLCA expression. When the resultant cell lines were injected subcutaneously into BALB/c mice, both produced tumors over the course of 3 months in two out of three mice injected, while non-selected HC11 cells failed to produce tumors. Again, loss of CLCA expression correlated with tumorigenicity.
Reduced Survival and Aberrant Morphology of Cells Transfected with mCLCA1 or mCLCA2-The loss of expression of both mCLCA1 and mCLCA2 during tumorigenesis suggested that either might inhibit survival of tumor cells. To test this idea, the cDNAs were cloned into pcDNA3.1 Zeo and transfected into low-passage HEK293 cells in a standard colony survival assay, used previously to demonstrate the growth inhibitory properties of tumor suppressors such as p16 and p19 (15). Co-transfection with a GFP-expressing plasmid provided a control for transfection frequency, which was about 20% in all dishes. However, when cells were subjected to selection with zeocin, although the same density of single colonies formed initially, nearly all of those transfected with mCLCA1 or mCLCA2 had died by 4 weeks of selection (Fig. 5A). A similar result was obtained when human breast cancer cell line MCF7 was transfected with mCLCA2 (Fig. 5A). Ultimately, only two mCLCA2-expressing HEK colonies survived; upon expansion, they grew slowly and cultures accumulated enlarged, multinucleated cells (Fig. 6, A-D). These results, together with the dramatic and divergent modulation of mCLCA1 and mCLCA2 expression in HC11, suggested that HC11 cells might also be sensitive to ectopic overexpression of mCLCA1 and mCLCA2. Indeed, while transfection with vector alone produced rapidly growing colonies containing cells of similar morphology to the starting population, overexpression of mCLCA1 or mCLCA2 resulted in few colonies (Fig. 5A), consisting of greatly enlarged, slowly growing cells (Fig. 6, E-G).
Because of the long-term nature of the antibiotic selection and low number of surviving colonies, we could not be sure whether giant cell formation was a direct effect of mCLCA2 over-expression or an adaptation that allowed the cell to tolerate mCLCA2 overexpression. Transfection of mCLCA2 under control of a regulated promoter failed to produce mCLCA2expressing colonies. Therefore a retrovirus was constructed that allowed efficient introduction of mCLCA2 into HC11 cells along with a GFP marker (mCLCA2-Pinco). One week after infection, cells were diluted and replated to allow single colony formation. Multinucleated giant cells were observed at a high frequency among mCLCA2-Pinco-infected cells but not among cells infected with the Pinco vector (Fig. 7). The largest cells exhibited up to 16 distinct but interconnected nuclear bodies, while cells of intermediate size contained fewer nuclei, most having no apparent interconnections (Fig. 7, columns 2 and 3). When cells were subjected to serum and anchorage deprivation, mCLCA2-overexpressing cells had twice the rate of apoptosis of vector-infected cells (Fig. 5B). DISCUSSION We describe here two closely related chloride channels whose regulation in response to apoptotic stress and growth factors is nearly reciprocal, suggesting opposing functions for mCLCA1 and mCLCA2. Surprisingly, however, expression of both genes is down-regulated in mammary tumors and cells selected for apoptosis-resistance. This dual down-regulation is not fortuitous, as would be expected if both genes relied on the same transcription factor for example, because in one case splicing of the pre-mRNA is deranged while in the other case the entire message becomes undetectable, consistent with a transcription-initiation defect. Moreover, transfection of either cDNA into several cell lines inhibits cell survival, with mCLCA2 more potent than mCLCA1. Thus, mCLCA1 and mCLCA2 are probably similar in function, with mCLCA2 specializing in stress response. Perhaps mCLCA2 represents a recently evolved "activated" form of mCLCA1, much as v-Ras is an activated form of c-Ras (27). Evidence for this hypothesis awaits detailed electrophysiological studies of these channels in HC11 and transfected cells under physiological conditions.
The finding that mCLCA2 overexpression in HC11 mammary epithelial cells sensitizes the cells to anoikis indicates that mCLCA2 plays some role in this process and may explain why tumor cells cannot tolerate its expression. These conditions were selected because they are most relevant to mammary involution and early tumor progression, but also because HC11 cells are p53-deficient and induction of apoptosis by growth factor and anchorage deprivation has been shown to be p53-independent in HC11 (28,29). Moreover, the absence of functional p53 prevented the fair testing of other apoptotic stimuli such as hypoxia and DNA damage because these pathways depend upon p53 to varying degrees (29 -32). Evidence is steadily accumulating that plasma membrane Cl Ϫ channels are critical effectors in the apoptotic pathway; activation of chloride currents in response to apoptogens has been observed in cells as diverse as human lymphocytes and frog oocytes (10,33). How might Cl Ϫ channels potentiate apoptosis? One possibility is regulation of intracellular pH (pH i ). Some Cl Ϫ channels can reduce pH i directly by also conducting bicarbonate out of the cell, as shown for CFTR in 3T3 and mammary epithelial cells (34) and the outwardly rectifying chloride channel in several cell types (35,36). Alternatively, by releasing Cl Ϫ , the channel might favor bicarbonate efflux via the Cl Ϫ /HCO 3Ϫ exchanger. Such an association between Cl Ϫ current and pH i reduction was observed in lymphocytes where ligation of the Fas receptor triggered an outwardly rectified Cl Ϫ current and a decrease in pH i , followed by apoptosis (10). Blockade of the channel prevented the decrease in pH i and inhibited apoptosis. Other workers have shown that preventing reduction of pH i is sufficient to prevent apoptosis (37). Similarly, mutation of CFTR in mammary epithelial cells inhibited intracellular acidification and apoptosis in response to cycloheximide treatment, while artificial reduction of pH i in the mutant cells restored sensitivity to an apoptotic stimulus (11). One consequence of intracellular acidification is the activation of a special class of endonucleases responsible for digesting chromatin into small fragments that are subsequently packaged into apoptotic bodies, a hallmark of apoptosis (30,38,39). It will be interesting to determine whether ectopic expression of CLCA genes reduces pH i .
The morphological aberrations associated with CLCA overexpression suggest an additional or alternative mode of tumor growth suppression. In mammary epithelial cells, the formation of multinuclear giants is associated with mutation or perturbed regulation of cell cycle regulators such as cyclins and p21 (40). For example, underexpression of BRCA1 partner BARD1 or overexpression of MDM2, both of which lead to lower levels of S phase inhibitors, produce identical morphological changes to those observed here (41,42). This phenotype resembles "premature senescence". Senescence is thought to be a mode of tumor suppression parallel to apoptosis and is typified by G 1 arrest in response to such diverse stimuli as telomere erosion, oxidative damage, and ectopic expression of p21 or activated Ras in a normal cell background (43)(44)(45). The potentiation of either phenotype by mCLCA2 is reminiscent of the effect of JunD deletion in mouse fibroblasts. Loss of JunD causes an accumulation of giant cells as well as greatly sensitizing cells to stress-induced apoptosis (46). On the other hand, two observations in this work suggest that CLCA expression may play an inhibitory role in the normal cell cycle. First, mCLCA1 is widely expressed in vivo and in HC11 cells in normal culture. Yet ectopic overexpression clearly limits cell proliferation in vitro. Indeed, a number of recent studies link plasma membrane ion channels to cell proliferation and cycle progression. For example, induction of potassium channels was found to stimulate proliferation of hepatocytes, while channel blockade reduced proliferation of these cells (47), T lymphocytes (48), melanoma (49), breast cancer cells (50), keratinocytes (51), and glial cells (52). The last study linked these effects to changes in pH i . Sustained blockage of the K ϩ channel in MCF7 breast cancer cells caused arrest in G 1 and eventual apoptosis (53). The activity of Cl Ϫ channels has also been found to vary with the cell cycle (54,55). In glioma cells and lymphocytes, Cl Ϫ current peaks in G 1 and is lowest during S phase (54,56). However, in contrast to K ϩ channels, blockade of Cl Ϫ channels was found to enhance proliferation of several cell types, including lymphocytes (57), Schwann cells (58), and keratinocytes (51), implying that at least one Cl Ϫ channel can retard the cell cycle. The mechanism by which these effects are manifested is unknown. However, a biochemical and molecular investigation of cell cycle determinants in CLCA-overexpressing cells may provide the answer. If calcium-activated chloride channels do somehow limit cell cycle transit as these data suggest, their loss would afford a significant growth advantage to tumor cells. It is tempting to speculate that these channels are part of the anti-neoplasia policing mechanism of mammary epithelial cells. In view of the widespread expression of mCLCA1 and mCLCA2 and the existence of human homologs, one of which is known to be down-regulated in breast cancer (9), a more general relevance to cancer is likely.